Advancing Compressive Strength Prediction in Self-Compacting Concrete via Soft Computing: A Robust Modeling Approach

Abstract

Self-Compacting Concrete (SCC) is a unique type of concrete that can flow and fill spaces without the need for vibrating compaction, resulting in a dense and uniform material. This article focuses on predicting the compressive strength of SCC using Artificial Neural Networks. Specifically, the study employs multilayer perceptrons with back-propagation learning algorithms, which are commonly used in various problem-solving scenarios. The study covers essential components such as structure, algorithm, data preprocessing, over-fitting prevention, and sensitivity analysis in MLPs. The input variables considered in the research include cement, limestone powder, fly ash, ground granulated blast furnace slag, silica fume, rice husk ash, coarse aggregate, fine aggregate, water, super-plasticizer, and viscosity-modifying admixtures. The target variable is the compressive strength. Through a sensitivity analysis, the study evaluates the relative importance of each parameter. The results demonstrate that the AI-based model accurately predicts the compressive strength of self-compacting concrete.

Keywords

• self-compacting concrete
• compressive strength prediction
• artificial neural network
• sensitivity analysis
• robust approach